Disordered Systems and Neural Networks
See recent articles
Showing new listings for Monday, 6 January 2025
- [1] arXiv:2501.01666 [pdf, html, other]
-
Title: Topological Anderson insulators by latent symmetryComments: 9 pages, 5 figures, Comments are welcomeSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Mesoscale and Nanoscale Physics (cond-mat.mes-hall)
Topological Anderson insulators represent a class of disorder-induced, nontrivial topological states. In this study, we propose a feasible strategy to unveil and design the latent-symmetry protected topological Anderson insulators. By employing the isospectral reduction approach from graph theory, we reduce a family of the disordered multi-atomic chains to the disordered dimerized chain characterized by energy-dependent potentials and hoppings, which exhibits the chiral symmetry or inversion symmetry. According to the topological invariants, bulk polarization, and the divergence of localization length of the topological bound edge states in the reduced disordered system, the gapped and ungapped topological Anderson states with latent symmetry could be identified in the original disordered multi-atomic systems. The concept of topological Anderson insulating phases protected by the geometric symmetries and tenfold-way classification is thus extended to the various types of latent symmetry cases. This work paves the way for exploiting topological Anderson insulators in terms of latent symmetries.
- [2] arXiv:2501.01853 [pdf, other]
-
Title: A self-learning magnetic Hopfield neural network with intrinsic gradient descent adaptionChang Niu, Huanyu Zhang, Chuanlong Xu, Wenjie Hu, Yunzhuo Wu, Yu Wu, Yadi Wang, Tong Wu, Yi Zhu, Yinyan Zhu, Wenbin Wang, Yizheng Wu, Lifeng Yin, Jiang Xiao, Weichao Yu, Hangwen Guo, Jian ShenComments: 21 pages, 5 figuresJournal-ref: Proc. Natl. Acad. Sci. U.S.A. 121 (51) e2416294121,(2024)Subjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Applied Physics (physics.app-ph)
Physical neural networks using physical materials and devices to mimic synapses and neurons offer an energy-efficient way to implement artificial neural networks. Yet, training physical neural networks are difficult and heavily relies on external computing resources. An emerging concept to solve this issue is called physical self-learning that uses intrinsic physical parameters as trainable weights. Under external inputs (i.e. training data), training is achieved by the natural evolution of physical parameters that intrinsically adapt modern learning rules via autonomous physical process, eliminating the requirements on external computation this http URL, we demonstrate a real spintronic system that mimics Hopfield neural networks (HNN) and unsupervised learning is intrinsically performed via the evolution of physical process. Using magnetic texture defined conductance matrix as trainable weights, we illustrate that under external voltage inputs, the conductance matrix naturally evolves and adapts Oja's learning algorithm in a gradient descent manner. The self-learning HNN is scalable and can achieve associative memories on patterns with high similarities. The fast spin dynamics and reconfigurability of magnetic textures offer an advantageous platform towards efficient autonomous training directly in materials.
New submissions (showing 2 of 2 entries)
- [3] arXiv:2501.01823 (cross-list from cond-mat.stat-mech) [pdf, html, other]
-
Title: Projected ensemble in a system with conserved charges with local supportSubjects: Statistical Mechanics (cond-mat.stat-mech); Disordered Systems and Neural Networks (cond-mat.dis-nn); Quantum Physics (quant-ph)
The investigation of ergodicity or lack thereof in isolated quantum many-body systems has conventionally focused on the description of the reduced density matrices of local subsystems in the contexts of thermalization, integrability, and localization. Recent experimental capabilities to measure the full distribution of quantum states in Hilbert space and the emergence of specific state ensembles have extended this to questions of {\textit{deep thermalization}}, by introducing the notion of the {\textit{projected ensemble}} -- ensembles of pure states of a subsystem obtained by projective measurements on its complement. While previous work examined chaotic unitary circuits, Hamiltonian evolution, and systems with global conserved charges, we study the projected ensemble in systems where there are an extensive number of conserved charges all of which have (quasi)local support. We employ a strongly disordered quantum spin chain which shows many-body localized dynamics over long timescales as well as the $\ell$-bit model, a phenomenological archetype of a many-body localized system, with the charges being $1$-local in the latter. In particular, we discuss the dependence of the projected ensemble on the measurement basis. Starting with random direct product states, we find that the projected ensemble constructed from time-evolved states converges to a Scrooge ensemble at late times and in the large system limit except when the measurement operator is close to the conserved charges. This is in contrast to systems with global conserved charges where the ensemble varies continuously with the measurement basis. We relate these observations to the emergence of Porter-Thomas distribution in the probability distribution of bitstring measurement probabilities.
Cross submissions (showing 1 of 1 entries)
- [4] arXiv:2404.14916 (replaced) [pdf, html, other]
-
Title: Anomalous Polarization in One-dimensional Aperiodic InsulatorsComments: Accepted for publication in MDPI Condensed MatterSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Mesoscale and Nanoscale Physics (cond-mat.mes-hall); Strongly Correlated Electrons (cond-mat.str-el)
By implementing a charge pumping scheme for one-dimensional aperiodic chains, we confirm the existence of topological phases in these systems whenever their finite-size realizations admit inversion symmetry. These phases are usually characterized by an anomalous edge response as a result of the bulk-boundary correspondence. We show that these signatures are all present in various chains, each representing a different class of structural aperiodicity: the Fibonacci quasicrystal, the Tribonacci quasicrystal, and the Thue-Morse chain. More specifically, we calculate three quantities: the Berry phase of the crystalline approximation of the finite-size systems, the polarization response to an inifintesimal static and constant electric field in systems with open boundary conditions, and the degeneracy of the entanglement spectrum. We find that all of them provide signatures of a topologically nontrivial phase.
- [5] arXiv:2312.08919 (replaced) [pdf, html, other]
-
Title: Scaling and data collapse of two-dimensional random singlet state in magnetic fieldComments: 5+epsilon pages, 3 figuresSubjects: Strongly Correlated Electrons (cond-mat.str-el); Disordered Systems and Neural Networks (cond-mat.dis-nn); Statistical Mechanics (cond-mat.stat-mech)
Quenched randomness strongly affects properties of magnetic materials. Two-dimensional (2D) random singlet (RS) states emerge in random $J$-$Q$ model by destroying valence bond solid order with spatial randomness. We examine the 2D RS state in magnetic field with quantum Monte Carlo simulations. The magnetization and susceptibilities show power-law scaling with magnetic field at low temperature. Moreover, they show one-parameter scaling behavior with $B/T$, and the scaling functions are remarkably consistent with a phenomenological model of random spin pairs with a singular distribution of interactions. These universal scaling functions can be used to diagnose 2D RS states in experiments.
- [6] arXiv:2405.06494 (replaced) [pdf, html, other]
-
Title: Experimental identification of topological defects in 2D colloidal glassJournal-ref: Nature Communications 16, 55 (2025)Subjects: Soft Condensed Matter (cond-mat.soft); Disordered Systems and Neural Networks (cond-mat.dis-nn); Materials Science (cond-mat.mtrl-sci); Other Condensed Matter (cond-mat.other); Statistical Mechanics (cond-mat.stat-mech)
Topological defects are singularities within a field that cannot be removed by continuous transformations. The definition of these irregularities requires an ordered reference configuration, calling into question whether they exist in disordered materials, such as glasses. However, recent work suggests that well-defined topological defects emerge in the dynamics of glasses, even if they are not evident in the static configuration. In this study, we reveal the presence of topological defects in the vibrational eigenspace of a two-dimensional experimental colloidal glass. These defects strongly correlate with the vibrational features and spatially correlate with each other and structural "soft spots", more prone to plastic flow. This work experimentally confirms the existence of topological defects in disordered systems revealing the complex interplay between topology, disorder, and dynamics.